How can I rotate column titles in pyplot.table?

Question

I'm creating a table in matplotlib, but the table headers are long strings, and the table values are numbers with only a few digits. This leaves me with two bad options: either my table is much wider than necessary, or my headers overlap. To fix this, I'd like to rotate the table headings (possibly up to 90 degrees). In other words, I want to do this in python

Here's my simplified code right now:

import matplotlib, numpy
import matplotlib.pyplot as plt

data=numpy.array([[1, 2],[3,4]])
headings=['long heading 1', 'long heading 2']
fig=plt.figure(figsize=(5,2))
ax=fig.add_subplot(111, frameon=False, xticks=[], yticks=[])

the_table = plt.table(cellText=data, rowLabels=headings, colLabels=headings, colWidths=[0.3]*data.shape[1], loc='center') #0.3 for second image, 0.03 for first
#the_table.auto_set_font_size(False) #comment out for second image
#the_table.set_fontsize(10) #comment out for second image
the_table.scale(1, 1.6)
plt.show()

This produces either the squished image, or the super-wide image (both shown below). In my actual code, the table is ~30 x 30, so the cells can't be very wide. Does anybody know how to rotate the column titles to fix this spacing issue?

Answer 1

The following worked for me.

create the table

 fig = plt.figure( figsize=(pageWidthInInches, pageHeightInInches) )
    panel = plotUtils.createPanelSameSizeAsFig(fig)

    tablePanel = panel.table(
                        cellText=cellText
                        # ,rowLabels=rowLabels
                        # ,colLabels=colLabels
                        #,loc='center' # center table in panel, title is in center
                        #,loc='bottom' # center table in panel does not work well
                        ,loc='best'
                        #,rowColours=aedwip,
                        #,colColourslist=aedwip
                        ,cellColours= cellColors
                       )

    # get rid of bar chart axis and box 
    panel.get_xaxis().set_visible(False)
    panel.get_yaxis().set_visible(False)

    tablePanel.scale(1, 1.5)

    plt.box(on=None)
    panel.set_title(title)

Now add the col header. Note my table did not have row labels, you may have to tweak the startX position

tcell = table._cells[(0, 0)]
cellWidth = tcell.get_width()
startX  = tcell.get_x() - cellWidth
y = 0.99 #0.98 #0.975 #0.96 #1

headings = sampleDF.columns
for i in range(len(headings)): 
    heading = headings[i]
    x = startX + i * cellWidth
    
    panel.text(x, y, heading, horizontalalignment="left", 
              verticalalignment="baseline", rotation=45, fontsize=4) 
  

Answer 2

I just found an other solution:

for cell in table._cells:
    if cell[0] ==0:
        table._cells[cell].get_text().set_rotation(90)

The first loop is for going through all cells, the second is for picking the first row/header.

if cell[1] =- -1 

this would selecte the the first column, which you might want to rotate, too.

Then you can rotate the cell text by e.g. 90 °.

Answer 3

I figured it out. It's not pretty, but it works. I added two annotations for each column - the text, and a line to separate it from the next column header. I had to define some parameters that apply to the table and the fancy labels (width, height, col_width), and some parameters to make the fancy labels line up correctly. This solution worked fine on my 30x30 table.

import matplotlib, numpy
import matplotlib.pyplot as plt

width=5
height=3
col_width=.075

data=numpy.array([[1, 2,5],[3,4,7],[7,9,5]])
headings=['long heading 1', 'long heading 2', 'longish 3']
fig=plt.figure(figsize=(width,height))
ax=fig.add_subplot(111, frameon=False, xticks=[], yticks=[])

the_table = plt.table(cellText=data, rowLabels=headings, 
    colWidths=[col_width]*data.shape[1], loc='center') #remove colLabels
the_table.auto_set_font_size(False) 
the_table.set_fontsize(10) 
the_table.scale(1, 1.6)

#custom heading titles - new portion
hoffset=0.42 #find this number from trial and error
voffset=0.66 #find this number from trial and error
line_fac=0.98 #controls the length of the dividing line
count=0
for string in headings:
    ax.annotate('  '+string, xy=(hoffset+count*col_width,voffset),
        xycoords='axes fraction', ha='left', va='bottom', 
        rotation=45, size=10)

    #add a dividing line
    ax.annotate('', xy=(hoffset+(count+0.5)*col_width,voffset), 
        xytext=(hoffset+(count+0.5)*col_width+line_fac/width,voffset+line_fac/height),
        xycoords='axes fraction', arrowprops={'arrowstyle':'-'})

    count+=1

plt.show()